Re: optimization/2938: [Regression vs 2.95] Optimization with -O1causes near infinite time

Re: optimization/2938: [Regression vs 2.95] Optimization with -O1causes near infinite time

[Wolfgang Bangerth]

The following reply was made to PR optimization/2938; it has been noted by GNATS.

From: Wolfgang Bangerth <wolfgang.bangerth@iwr.uni-heidelberg.de>
To: mmitchel@gcc.gnu.org
Cc: gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org
Subject: Re: optimization/2938: [Regression vs 2.95] Optimization with -O1
 causes near infinite time
Date: Mon, 25 Jun 2001 14:59:57 +0200 (MET DST)

 > State-Changed-From-To: analyzed->closed
 >     Fixed.
 
 Dear Mark,
 thanks for looking into it. However, in the 3.0 release your fix seems to
 not be working: for c++615, which is kind of a small form of
 optimization/2938, I still get 1009 calls to Point::operator() (which is,
 as you said, not only a problem of optimization, but also a correctness
 issue as the operator is not marked constant). This is both on Sun
 Solaris2.7 and Linux.
 
 Optimization/2938 still takes infintely long. I double checked, but my
 gcc3 sources have your patch to the problem incorporated. Do you know what
 is going on?
 
 Regards
   Wolfgang
 
 var/gcc-bugs> c++ -v
   Reading specs from
   /home/people/wolf/Config/gcc-ss-sun/lib/gcc-lib/sparc-sun-solaris2.7/3.0/specs
   Configured with: ../gcc-3.0/configure
   --prefix=/home/people/wolf/Config/gcc-ss-sun --with-gnu-as
   --with-as=/usr/local/bin/as --with-gnu-ld --with-ld=/usr/local/bin/ld
   Thread model: posix
   gcc version 3.0
 var/gcc-bugs> c++ -O1 -S 0615+.cc
 var/gcc-bugs> cat 0615+.s | c++filt | grep operator | uniq
         call    Point::operator()(unsigned), 0
 var/gcc-bugs> cat 0615+.s | c++filt | grep operator | wc -l
    1009
 
 
 -------------------------------------------------------------------------
 Wolfgang Bangerth          email: wolfgang.bangerth@iwr.uni-heidelberg.de
                              www: http://gaia.iwr.uni-heidelberg.de/~wolf
 
 
 

Re: optimization/2938: [Regression vs 2.95] Optimization with -O1causes near infinite time

[Wolfgang Bangerth]

The following reply was made to PR optimization/2938; it has been noted by GNATS.

From: Wolfgang Bangerth <wolfgang.bangerth@iwr.uni-heidelberg.de>
To: Mark Mitchell <mark@codesourcery.com>
Cc: "mmitchel@gcc.gnu.org" <mmitchel@gcc.gnu.org>,
   "gcc-bugs@gcc.gnu.org" <gcc-bugs@gcc.gnu.org>,
   "gcc-gnats@gcc.gnu.org" <gcc-gnats@gcc.gnu.org>
Subject: Re: optimization/2938: [Regression vs 2.95] Optimization with -O1
 causes near infinite time
Date: Mon, 2 Jul 2001 16:21:11 +0200 (MET DST)

 Mark Mitchell wrote:
 > I have just checked in a patch that I think will really work this time.
 > Would you check that it fixes the problem for you, and then tell me
 > which PRs I can close?
 
 It indeed does this time :-) optimization/2938 is already closed, but
 c++/615 isn't, so please close that. (615 was the original report, 2938
 was kind of a re-post that showed the problem in a more pronounced way).
 
 Many thanks & best regards
   Wolfgang
 
 PS: Checked on Solaris2.7 and Linux.
 
 -------------------------------------------------------------------------
 Wolfgang Bangerth          email: wolfgang.bangerth@iwr.uni-heidelberg.de
                              www: http://gaia.iwr.uni-heidelberg.de/~wolf
 
 
 

Re: optimization/2938: [Regression vs 2.95] Optimization with -O1causes near infinite time

[Wolfgang Bangerth]

The following reply was made to PR optimization/2938; it has been noted by GNATS.

From: Wolfgang Bangerth <wolfgang.bangerth@iwr.uni-heidelberg.de>
To: Mark Mitchell <mark@codesourcery.com>
Cc: "gcc-bugs@gcc.gnu.org" <gcc-bugs@gcc.gnu.org>, gcc-gnats@gcc.gnu.org
Subject: Re: optimization/2938: [Regression vs 2.95] Optimization with -O1
 causes near infinite time
Date: Tue, 26 Jun 2001 08:42:01 +0200 (MET DST)

 > >> State-Changed-From-To: analyzed->closed
 > >>     Fixed.
 > >
 > > Dear Mark,
 > > thanks for looking into it.
 > 
 > I will investigate.
 
 You may want to use the appended tiny program for the regression test
 suite or for debugging. If it helps, I can as well file another bug report
 with a more suitable name (please propose one) so that c++/615 and
 optimization/2938 can be closed.
 
 Regards
   Wolfgang
 
 
 -----------------------------------------
 // compile with -O1
 #include <assert.h>
 
 // assert that foo() is only called once
 int count = 0;
 
 double foo () {
   count++;
   return 0;
 };
 
 double bar () {
   const double x[1] = { foo() };
   return x[0];
 };
 
 int main () 
 {
   bar();
   assert (count == 1);
 };
 
 -------------------------------------------------------------------------
 Wolfgang Bangerth          email: wolfgang.bangerth@iwr.uni-heidelberg.de
                              www: http://gaia.iwr.uni-heidelberg.de/~wolf